cm013 Exercise

suppressPackageStartupMessages(library(tidyverse))
library(gapminder)

Saving Graphs to File

• Don’t use the mouse
• Use ggsave for ggplot
  • Practice by saving the following plot to file:

ggplot(mtcars, aes(hp, wt)) + 
    geom_point()

#ggsave(FILENAME_HERE, PLOT_OBJECT_HERE)

• Base R way: print plots “to screen”, sandwiched between pdf()/jpeg()/png()… and dev.off().
• Vector vs. raster: Images are stored on your computer as either vector or raster.
  • Raster: an n by m grid of pixels, each with its own colour. jpeg, png, gif, bmp.
  • Vector: represented as shapes and lines. pdf, svg.
  • For tips: “10 tips for making your R graphics look their best”“.

Scales; Colour

Scale functions in ggplot2 take the form scale_[aesthetic]_[mapping]().

Let’s first focus on the following plot:

p_scales <- ggplot(gapminder, aes(gdpPercap, lifeExp)) +
     geom_point(aes(colour=pop), alpha=0.2)
p_scales + 
    scale_x_log10() +
    scale_colour_continuous(trans="log10")

1. Change the y-axis tick mark spacing to 10; change the colour spacing to include all powers of 10.

p_scales +
    scale_x_log10() +
    scale_colour_continuous(
        trans  = "log10", 
        breaks = 10^(1:10)
    ) +
    scale_y_continuous(breaks=1:10 * 10) #vector of 1 to 10, multiplied by 10 

#indicates what number we'd like to put tick marks for; if you're out of range, default is to exclude those tick marks
#continuous does not transform the axis at all

2. Specify scales::*_format in the labels argument of a scale function to do the following:
   • Change the x-axis labels to dollar format (use scales::dollar_format())
   • Change the colour labels to comma format (use scales::comma_format())

library(scales)
p_scales +
    scale_x_log10(labels=dollar_format()) +
    scale_colour_continuous(
        trans  = "log10", 
        breaks = 10^(1:10),
        labels = comma_format()
    ) +
    scale_y_continuous(breaks=10*(1:10))

3. Use RColorBrewer to change the colour scheme.
   • Notice the three different types of scales: sequential, diverging, and continuous. For discrete or categorical variables, use very different colours. -> continuous For a sequence of variables, use a gradient of colours. -> sequential For a middle neutral value with a scale of gradients moving away in each direction (i.e. temperature scale to hot and cold), -> diverging

## All palettes the come with RColorBrewer:
RColorBrewer::display.brewer.all()

p_scales +
    scale_x_log10(labels=dollar_format()) +
    scale_colour_distiller(  #another one is scale_colour_brewer (one is for continuous, other handles the other two types)
        trans   = "log10",
        breaks  = 10^(1:10),
        labels  = comma_format(),
        palette = "Greens" #get the name from looking at RColorBrewer overview
    ) +
    scale_y_continuous(breaks=10*(1:10))

4. Use viridis to change the colour to a colour-blind friendly scheme
   • Hint: add scale_colour_viridis_c (c stands for continuous; d discrete).
   • You can choose a palette with option.

p_scales +
    scale_x_log10(labels=dollar_format()) +
    scale_colour_viridis_c(
        trans   = "log10",
        breaks  = 10^(1:10),
        labels  = comma_format()
    ) +
    scale_y_continuous(breaks=10*(1:10))

Theming

Changing the look of a graphic can be achieved through the theme() layer.

There are “complete themes” that come with ggplot2, my favourite being theme_bw (I’ve grown tired of the default gray background, so theme_bw is refreshing).

1. Change the theme of the following plot to theme_bw():

ggplot(iris, aes(Sepal.Width, Sepal.Length)) +
     facet_wrap(~ Species) +
     geom_point() +
     labs(x = "Sepal Width",
          y = "Sepal Length",
          title = "Sepal sizes of three plant species") +
    theme_bw() #many possibilities (see tab over)

2. Then, change font size of axis labels, and the strip background colour. Others?

ggplot(iris, aes(Sepal.Width, Sepal.Length)) +
     facet_wrap(~ Species) +
     geom_point() +
     labs(x = "Sepal Width",
          y = "Sepal Length",
          title = "Sepal sizes of three plant species") +
    theme_bw() +
    theme(axis.text = element_text(size = 16),
          strip.background = element_rect(fill = "orange"),
          panel.background = element_rect(fill = "blue"))

Plotly

Consider the following plot:

(p <- gapminder %>% 
     filter(continent != "Oceania") %>% 
     ggplot(aes(gdpPercap, lifeExp)) +
     geom_point(aes(colour=pop), alpha=0.2) +
     scale_x_log10(labels=dollar_format()) +
     scale_colour_distiller(
         trans   = "log10",
         breaks  = 10^(1:10),
         labels  = comma_format(),
         palette = "Greens"
     ) +
     facet_wrap(~ continent) +
     scale_y_continuous(breaks=10*(1:10)) +
     theme_bw())

1. Convert it to a plotly object by applying the ggplotly() function:

#you can make a ggplot object and convert it to plotly (dont need to know plotly syntax)
#install.packages("plotly")
library(plotly)

Attaching package: 㤼㸱plotly㤼㸲

The following object is masked from 㤼㸱package:ggplot2㤼㸲:

    last_plot

The following object is masked from 㤼㸱package:stats㤼㸲:

    filter

The following object is masked from 㤼㸱package:graphics㤼㸲:

    layout

ggplotly(p)

2. You can save a plotly graph locally as an html file. Try saving the above:
   • NOTE: plotly graphs don’t seem to show up in Rmd notebooks, but they do with Rmd documents.

p %>% 
    ggplotly() %>% 
    htmlwidgets::saveWidget("LOCATION_GOES_HERE")

3. Run this code to see the json format underneath:

p %>% 
    ggplotly() %>% 
    plotly_json()

4. Check out code to make a plotly object from scratch using plot_ly() – scatterplot of gdpPercap vs lifeExp.
   • Check out the cheat sheet.

plot_ly(gapminder, 
        x = ~gdpPercap, #tilda means make a formula with this variable as your formula
        y = ~lifeExp, 
        type = "scatter",
        mode = "markers",
        opacity = 0.2) %>% 
    layout(xaxis = list(type = "log"))

5. Add population to form a z-axis for a 3D plot:

plot_ly(gapminder, 
        x = ~gdpPercap, 
        y = ~lifeExp, 
        z = ~pop,
        type = "scatter3d",
        mode = "markers",
        opacity = 0.2)

can use gg animate function to animate plots

Why do we make plots? EDA (exploratory data analysis); drawing conclusions from plots -> endpoint - we’ve looked at tooling with plots -> intermediate - grammar and theme are the two aspects that we make decisions about when producing a plot -> intermediate - effectiveness; what information is contained in the data? -> starting point - see jenny’s animated gif for simplfying data -> does it facilitate comparisons or reveal trends? - see Tamara Munsener’s book for psychology of colour effects on perception - don’t use pie charts! can be difficult to compare categories (hard to perceive angular distances vs linear, as in a bar plot) - when you look at your plot, ask yourself what trends can you see; what is hidden; how can you reveal it? - impossible to make a perfect plot (this is why we have figure captions) - publication quality: - no colour - resolution/aspect ratio/size - font size - sans serif - labels, units - consistency - inluding zero is important for bar charts (e.g.) bc we need to visually gauge the entire area of the bar for comparison